Figure 1.
The instrument for FLIP measurements.
(A) The imaging LED and bleaching laser are combined to excite GFP or GFP-labeled proteins in the cell (ribbon structure of PGK at the top right). The LED illuminates the whole cell evenly for imaging. The laser is focused to a small intense spot (see Methods) to locally photobleach the fluorescent protein. (B) The LED and laser are controlled to turn on alternately every 10 seconds. Snapshots are taken with only the LED on to record the progress of the fluorescence intensity decay in the cell without saturating the camera.
Figure 2.
Snapshots of the GFP fluorescence intensity at three of the 19 time steps sampled during bleaching in U2OS cells.
Left column: FLIP data in true color; the small bleaching spot and nucleus are excluded from analysis. Middle three columns: FLIP data in false color for better contrast (scale bar at right), 3-D diffusion model fit, and fit residual. Rightmost three columns: Same FLIP data in false color normalized to the fluorescence intensity distribution at t = 0, 2-D diffusion model fit, and fit residual.
Figure 3.
Comparison of experimental data with 2-D normal and anomalous diffusion models at the two locations in the cell (A) and (B).
The experimental fluorescence intensity (black circles) initially decays faster, and subsequently slower than the best normal diffusion model fits (blue curves). Only anomalous diffusion with α<1 (red curves) correctly simulates the observed data. The vertical dashed lines at 85 s indicate where the short time (horizontal axis in Figure 4) and long time (vertical axis in Figure 4) residuals were calculated. The normal diffusion model tends to have a negative residual at short time and a positive residual at long time, whereas the anomalous diffusion model has much smaller residuals.
Figure 4.
The distribution of the residuals of all the pixels in the cell.
(A–E) Single and multi-domain models of normal and anomalous diffusion simulations. The pixels represent the residual of a single pixel in the cell at t≤90 s (x-axis) and t≥100 s (y-axis). Normal diffusion simulations (A,C) have systematic deviation from experiment results, visualized by an offset in the residual graph. In the anomalous diffusion simulations (B,D), the multi-domain model results in smaller overall residuals than the single domain model. (E) The illustration of the three domains in the cell calculated by the k-means clustering algorithm.
Figure 5.
Comparison of the diffusion coefficients of the GFP, ltPGK-GFP and ltPGK-FRET measured from 22°C to 37°C.
All proteins diffuse faster at higher temperatures while folded. The “lt” proteins show accelerated diffusion followed by a turnaround at T near Tm. Global model fits are to equation 1 (solid thick lines) and equation 2 (dotted thick lines).
Table 1.
2-D model fits for GFP diffusion inside the cell shown in Figures 2-4.
Table 2.
Mass and melting temperature of the main proteins in this study, as well as number of cells N measured by FLIP and standard deviation σ (in µm2/s) of the average measured diffusion coefficients (shown in Figure 5) among the N cells.
Figure 6.
Donor over acceptor (D/A) ratios for protein unfolding and protein-protein interaction.
PGK-FRET unfolding (gray) has a midpoint at approximately 42°C for the U2OS cell data shown here, unfolding begins at 37°C. ltPGK-FRET starts interacting with Hsp70-mCherry extensively above 35°C, whereas htPGK-FRET simply continues the room temperature linear trend up to 45°C.